eSpeak NG is an open source speech synthesizer that supports more than hundred languages and accents.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

mbrowrap.cpp 13KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613
  1. /*
  2. * mbrowrap -- A wrapper library around the mbrola binary
  3. * providing a subset of the API from the Windows mbrola DLL.
  4. *
  5. * Copyright (C) 2010 by Nicolas Pitre <[email protected]>
  6. *
  7. * This program is free software; you can redistribute it and/or modify
  8. * it under the terms of the GNU General Public License as published by
  9. * the Free Software Foundation; either version 3 of the License, or
  10. * (at your option) any later version.
  11. *
  12. * This program is distributed in the hope that it will be useful,
  13. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  15. * GNU General Public License for more details.
  16. */
  17. #include "speech.h"
  18. #ifdef INCLUDE_MBROLA
  19. #include <stdarg.h>
  20. #include <stdio.h>
  21. #include <unistd.h>
  22. #include <string.h>
  23. #include <stdlib.h>
  24. #include <signal.h>
  25. #include <fcntl.h>
  26. #include <poll.h>
  27. #include <errno.h>
  28. #include <sys/types.h>
  29. #include <sys/wait.h>
  30. #include "mbrowrap.h"
  31. /*
  32. * mbrola instance parameters
  33. */
  34. enum mbr_state {
  35. MBR_INACTIVE = 0,
  36. MBR_IDLE,
  37. MBR_NEWDATA,
  38. MBR_AUDIO,
  39. MBR_WEDGED
  40. };
  41. static enum mbr_state mbr_state;
  42. static char *mbr_voice_path;
  43. static int mbr_cmd_fd, mbr_audio_fd, mbr_error_fd, mbr_proc_stat;
  44. static pid_t mbr_pid;
  45. static int mbr_samplerate;
  46. static float mbr_volume = 1.0;
  47. static char mbr_errorbuf[160];
  48. struct datablock {
  49. struct datablock *next;
  50. int done;
  51. int size;
  52. char buffer[1]; /* 1 or more, dynamically allocated */
  53. };
  54. static struct datablock *mbr_pending_data_head, *mbr_pending_data_tail;
  55. /*
  56. * Private support code.
  57. */
  58. static void log(const char *msg, ...)
  59. {
  60. va_list params;
  61. va_start(params, msg);
  62. vfprintf(stderr, msg, params);
  63. fputc('\n', stderr);
  64. va_end(params);
  65. }
  66. static void err(const char *errmsg, ...)
  67. {
  68. va_list params;
  69. va_start(params, errmsg);
  70. vsnprintf(mbr_errorbuf, sizeof(mbr_errorbuf), errmsg, params);
  71. va_end(params);
  72. log("mbrowrap error: %s", mbr_errorbuf);
  73. }
  74. static int create_pipes(int p1[2], int p2[2], int p3[2])
  75. {
  76. int error;
  77. if (pipe(p1) != -1) {
  78. if (pipe(p2) != -1) {
  79. if (pipe(p3) != -1) {
  80. return 0;
  81. } else
  82. error = errno;
  83. close(p2[0]);
  84. close(p2[1]);
  85. } else
  86. error = errno;
  87. close(p1[0]);
  88. close(p1[1]);
  89. } else
  90. error = errno;
  91. err("pipe(): %s", strerror(error));
  92. return -1;
  93. }
  94. static void close_pipes(int p1[2], int p2[2], int p3[2])
  95. {
  96. close(p1[0]);
  97. close(p1[1]);
  98. close(p2[0]);
  99. close(p2[1]);
  100. close(p3[0]);
  101. close(p3[1]);
  102. }
  103. static int start_mbrola(const char *voice_path)
  104. {
  105. int error, p_stdin[2], p_stdout[2], p_stderr[2];
  106. ssize_t written;
  107. char charbuf[20];
  108. if (mbr_state != MBR_INACTIVE) {
  109. err("mbrola init request when already initialized");
  110. return -1;
  111. }
  112. error = create_pipes(p_stdin, p_stdout, p_stderr);
  113. if (error)
  114. return -1;
  115. mbr_pid = fork();
  116. if (mbr_pid == -1) {
  117. error = errno;
  118. close_pipes(p_stdin, p_stdout, p_stderr);
  119. err("fork(): %s", strerror(error));
  120. return -1;
  121. }
  122. if (mbr_pid == 0) {
  123. int i;
  124. if (dup2(p_stdin[0], 0) == -1 ||
  125. dup2(p_stdout[1], 1) == -1 ||
  126. dup2(p_stderr[1], 2) == -1) {
  127. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  128. "dup2(): %s\n", strerror(errno));
  129. written = write(p_stderr[1], mbr_errorbuf, strlen(mbr_errorbuf));
  130. (void)written; // suppress 'variable not used' warning
  131. _exit(1);
  132. }
  133. for (i = p_stderr[1]; i > 2; i--)
  134. close(i);
  135. signal(SIGHUP, SIG_IGN);
  136. signal(SIGINT, SIG_IGN);
  137. signal(SIGQUIT, SIG_IGN);
  138. signal(SIGTERM, SIG_IGN);
  139. snprintf(charbuf, sizeof(charbuf), "%g", mbr_volume);
  140. execlp("mbrola", "mbrola", "-e", "-v", charbuf,
  141. voice_path, "-", "-.wav", (char *)NULL);
  142. /* if execution reaches this point then the exec() failed */
  143. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  144. "mbrola: %s\n", strerror(errno));
  145. written = write(2, mbr_errorbuf, strlen(mbr_errorbuf));
  146. (void)written; // suppress 'variable not used' warning
  147. _exit(1);
  148. }
  149. snprintf(charbuf, sizeof(charbuf), "/proc/%d/stat", mbr_pid);
  150. mbr_proc_stat = open(charbuf, O_RDONLY);
  151. if (mbr_proc_stat == -1) {
  152. error = errno;
  153. close_pipes(p_stdin, p_stdout, p_stderr);
  154. waitpid(mbr_pid, NULL, 0);
  155. mbr_pid = 0;
  156. err("/proc is unaccessible: %s", strerror(error));
  157. return -1;
  158. }
  159. signal(SIGPIPE, SIG_IGN);
  160. if (fcntl(p_stdin[1], F_SETFL, O_NONBLOCK) == -1 ||
  161. fcntl(p_stdout[0], F_SETFL, O_NONBLOCK) == -1 ||
  162. fcntl(p_stderr[0], F_SETFL, O_NONBLOCK) == -1) {
  163. error = errno;
  164. close_pipes(p_stdin, p_stdout, p_stderr);
  165. waitpid(mbr_pid, NULL, 0);
  166. mbr_pid = 0;
  167. err("fcntl(): %s", strerror(error));
  168. return -1;
  169. }
  170. mbr_cmd_fd = p_stdin[1];
  171. mbr_audio_fd = p_stdout[0];
  172. mbr_error_fd = p_stderr[0];
  173. close(p_stdin[0]);
  174. close(p_stdout[1]);
  175. close(p_stderr[1]);
  176. mbr_state = MBR_IDLE;
  177. return 0;
  178. }
  179. static void stop_mbrola(void)
  180. {
  181. if (mbr_state == MBR_INACTIVE)
  182. return;
  183. close(mbr_proc_stat);
  184. close(mbr_cmd_fd);
  185. close(mbr_audio_fd);
  186. close(mbr_error_fd);
  187. if (mbr_pid) {
  188. kill(mbr_pid, SIGTERM);
  189. waitpid(mbr_pid, NULL, 0);
  190. mbr_pid = 0;
  191. }
  192. mbr_state = MBR_INACTIVE;
  193. }
  194. static void free_pending_data(void)
  195. {
  196. struct datablock *p, *head = mbr_pending_data_head;
  197. while (head) {
  198. p = head;
  199. head = head->next;
  200. free(p);
  201. }
  202. mbr_pending_data_head = NULL;
  203. mbr_pending_data_tail = NULL;
  204. }
  205. static int mbrola_died(void)
  206. {
  207. pid_t pid;
  208. int status, len;
  209. const char *msg;
  210. char msgbuf[80];
  211. pid = waitpid(mbr_pid, &status, WNOHANG);
  212. if (!pid) {
  213. msg = "mbrola closed stderr and did not exit";
  214. } else if (pid != mbr_pid) {
  215. msg = "waitpid() is confused";
  216. } else {
  217. mbr_pid = 0;
  218. if (WIFSIGNALED(status)) {
  219. int sig = WTERMSIG(status);
  220. snprintf(msgbuf, sizeof(msgbuf),
  221. "mbrola died by signal %d", sig);
  222. msg = msgbuf;
  223. } else if (WIFEXITED(status)) {
  224. int exst = WEXITSTATUS(status);
  225. snprintf(msgbuf, sizeof(msgbuf),
  226. "mbrola exited with status %d", exst);
  227. msg = msgbuf;
  228. } else {
  229. msg = "mbrola died and wait status is weird";
  230. }
  231. }
  232. log("mbrowrap error: %s", msg);
  233. len = strlen(mbr_errorbuf);
  234. if (!len)
  235. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf), "%s", msg);
  236. else
  237. snprintf(mbr_errorbuf + len, sizeof(mbr_errorbuf) - len,
  238. ", (%s)", msg);
  239. return -1;
  240. }
  241. static int mbrola_has_errors(void)
  242. {
  243. int result;
  244. char buffer[256];
  245. char *buf_ptr, *lf;
  246. buf_ptr = buffer;
  247. for (;;) {
  248. result = read(mbr_error_fd, buf_ptr,
  249. sizeof(buffer) - (buf_ptr - buffer) - 1);
  250. if (result == -1) {
  251. if (errno == EAGAIN)
  252. return 0;
  253. err("read(error): %s", strerror(errno));
  254. return -1;
  255. }
  256. if (result == 0) {
  257. /* EOF on stderr, assume mbrola died. */
  258. return mbrola_died();
  259. }
  260. buf_ptr[result] = 0;
  261. for (; (lf = strchr(buf_ptr, '\n')); buf_ptr = lf + 1) {
  262. /* inhibit the reset signal messages */
  263. if (strncmp(buf_ptr, "Got a reset signal", 18) == 0 ||
  264. strncmp(buf_ptr, "Input Flush Signal", 18) == 0)
  265. continue;
  266. *lf = 0;
  267. log("mbrola: %s", buf_ptr);
  268. /* is this the last line? */
  269. if (lf == &buf_ptr[result - 1]) {
  270. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  271. "%s", buf_ptr);
  272. /* don't consider this fatal at this point */
  273. return 0;
  274. }
  275. }
  276. memmove(buffer, buf_ptr, result);
  277. buf_ptr = buffer + result;
  278. }
  279. }
  280. static int send_to_mbrola(const char *cmd)
  281. {
  282. ssize_t result;
  283. int len;
  284. if (!mbr_pid)
  285. return -1;
  286. len = strlen(cmd);
  287. result = write(mbr_cmd_fd, cmd, len);
  288. if (result == -1) {
  289. int error = errno;
  290. if (error == EPIPE && mbrola_has_errors()) {
  291. return -1;
  292. } else if (error == EAGAIN) {
  293. result = 0;
  294. } else {
  295. err("write(): %s", strerror(error));
  296. return -1;
  297. }
  298. }
  299. if (result != len) {
  300. struct datablock *data;
  301. data = (struct datablock *)malloc(sizeof(*data) + len - result);
  302. if (data) {
  303. data->next = NULL;
  304. data->done = 0;
  305. data->size = len - result;
  306. memcpy(data->buffer, cmd + result, len - result);
  307. result = len;
  308. if (!mbr_pending_data_head)
  309. mbr_pending_data_head = data;
  310. else
  311. mbr_pending_data_tail->next = data;
  312. mbr_pending_data_tail = data;
  313. }
  314. }
  315. return result;
  316. }
  317. static int mbrola_is_idle(void)
  318. {
  319. char *p;
  320. char buffer[20]; /* looking for "12345 (mbrola) S" so 20 is plenty*/
  321. /* look in /proc to determine if mbrola is still running or sleeping */
  322. if (lseek(mbr_proc_stat, 0, SEEK_SET) != 0)
  323. return 0;
  324. if (read(mbr_proc_stat, buffer, sizeof(buffer)) != sizeof(buffer))
  325. return 0;
  326. p = (char *)memchr(buffer, ')', sizeof(buffer));
  327. if (!p || (unsigned)(p - buffer) >= sizeof(buffer) - 2)
  328. return 0;
  329. return (p[1] == ' ' && p[2] == 'S');
  330. }
  331. static ssize_t receive_from_mbrola(void *buffer, size_t bufsize)
  332. {
  333. int result, wait = 1;
  334. size_t cursize = 0;
  335. if (!mbr_pid)
  336. return -1;
  337. do {
  338. struct pollfd pollfd[3];
  339. nfds_t nfds = 0;
  340. int idle;
  341. pollfd[0].fd = mbr_audio_fd;
  342. pollfd[0].events = POLLIN;
  343. nfds++;
  344. pollfd[1].fd = mbr_error_fd;
  345. pollfd[1].events = POLLIN;
  346. nfds++;
  347. if (mbr_pending_data_head) {
  348. pollfd[2].fd = mbr_cmd_fd;
  349. pollfd[2].events = POLLOUT;
  350. nfds++;
  351. }
  352. idle = mbrola_is_idle();
  353. result = poll(pollfd, nfds, idle ? 0 : wait);
  354. if (result == -1) {
  355. err("poll(): %s", strerror(errno));
  356. return -1;
  357. }
  358. if (result == 0) {
  359. if (idle) {
  360. mbr_state = MBR_IDLE;
  361. break;
  362. } else {
  363. if (wait >= 5000 * (4-1)/4) {
  364. mbr_state = MBR_WEDGED;
  365. err("mbrola process is stalled");
  366. break;
  367. } else {
  368. wait *= 4;
  369. continue;
  370. }
  371. }
  372. }
  373. wait = 1;
  374. if (pollfd[1].revents && mbrola_has_errors())
  375. return -1;
  376. if (mbr_pending_data_head && pollfd[2].revents) {
  377. struct datablock *head = mbr_pending_data_head;
  378. char *data = head->buffer + head->done;
  379. int left = head->size - head->done;
  380. result = write(mbr_cmd_fd, data, left);
  381. if (result == -1) {
  382. int error = errno;
  383. if (error == EPIPE && mbrola_has_errors())
  384. return -1;
  385. err("write(): %s", strerror(error));
  386. return -1;
  387. }
  388. if (result != left) {
  389. head->done += result;
  390. } else {
  391. mbr_pending_data_head = head->next;
  392. free(head);
  393. if (!mbr_pending_data_head)
  394. mbr_pending_data_tail = NULL;
  395. else
  396. continue;
  397. }
  398. }
  399. if (pollfd[0].revents) {
  400. char *curpos = (char *)buffer + cursize;
  401. size_t space = bufsize - cursize;
  402. ssize_t obtained = read(mbr_audio_fd, curpos, space);
  403. if (obtained == -1) {
  404. err("read(): %s", strerror(errno));
  405. return -1;
  406. }
  407. cursize += obtained;
  408. mbr_state = MBR_AUDIO;
  409. }
  410. } while (cursize < bufsize);
  411. return cursize;
  412. }
  413. /*
  414. * API functions.
  415. */
  416. int init_MBR(const char *voice_path)
  417. {
  418. int error, result;
  419. unsigned char wavhdr[45];
  420. error = start_mbrola(voice_path);
  421. if (error)
  422. return -1;
  423. result = send_to_mbrola("#\n");
  424. if (result != 2) {
  425. stop_mbrola();
  426. return -1;
  427. }
  428. /* we should actually be getting only 44 bytes */
  429. result = receive_from_mbrola(wavhdr, 45);
  430. if (result != 44) {
  431. if (result >= 0)
  432. err("unable to get .wav header from mbrola");
  433. stop_mbrola();
  434. return -1;
  435. }
  436. /* parse wavhdr to get mbrola voice samplerate */
  437. if (memcmp(wavhdr, "RIFF", 4) != 0 ||
  438. memcmp(wavhdr+8, "WAVEfmt ", 8) != 0) {
  439. err("mbrola did not return a .wav header");
  440. stop_mbrola();
  441. return -1;
  442. }
  443. mbr_samplerate = wavhdr[24] + (wavhdr[25]<<8) +
  444. (wavhdr[26]<<16) + (wavhdr[27]<<24);
  445. //log("mbrowrap: voice samplerate = %d", mbr_samplerate);
  446. /* remember the voice path for setVolumeRatio_MBR() */
  447. if (mbr_voice_path != voice_path) {
  448. free(mbr_voice_path);
  449. mbr_voice_path = strdup(voice_path);
  450. }
  451. return 0;
  452. }
  453. void close_MBR(void)
  454. {
  455. stop_mbrola();
  456. free_pending_data();
  457. free(mbr_voice_path);
  458. mbr_voice_path = NULL;
  459. mbr_volume = 1.0;
  460. }
  461. int reset_MBR()
  462. {
  463. int result, success = 1;
  464. char dummybuf[4096];
  465. if (mbr_state == MBR_IDLE)
  466. return 1;
  467. if (!mbr_pid)
  468. return 0;
  469. if (kill(mbr_pid, SIGUSR1) == -1)
  470. success = 0;
  471. free_pending_data();
  472. result = write(mbr_cmd_fd, "\n#\n", 3);
  473. if (result != 3)
  474. success = 0;
  475. do {
  476. result = read(mbr_audio_fd, dummybuf, sizeof(dummybuf));
  477. } while (result > 0);
  478. if (result != -1 || errno != EAGAIN)
  479. success = 0;
  480. if (!mbrola_has_errors() && success)
  481. mbr_state = MBR_IDLE;
  482. return success;
  483. }
  484. int read_MBR(void *buffer, int nb_samples)
  485. {
  486. int result = receive_from_mbrola(buffer, nb_samples * 2);
  487. if (result > 0)
  488. result /= 2;
  489. return result;
  490. }
  491. int write_MBR(const char *data)
  492. {
  493. mbr_state = MBR_NEWDATA;
  494. return send_to_mbrola(data);
  495. }
  496. int flush_MBR(void)
  497. {
  498. return send_to_mbrola("\n#\n") == 3;
  499. }
  500. int getFreq_MBR(void)
  501. {
  502. return mbr_samplerate;
  503. }
  504. void setVolumeRatio_MBR(float value)
  505. {
  506. if (value == mbr_volume)
  507. return;
  508. mbr_volume = value;
  509. if (mbr_state != MBR_IDLE)
  510. return;
  511. /*
  512. * We have no choice but to kill and restart mbrola with
  513. * the new argument here.
  514. */
  515. stop_mbrola();
  516. init_MBR(mbr_voice_path);
  517. }
  518. int lastErrorStr_MBR(char *buffer, int bufsize)
  519. {
  520. int result;
  521. if (mbr_pid)
  522. mbrola_has_errors();
  523. result = snprintf(buffer, bufsize, "%s", mbr_errorbuf);
  524. return result >= bufsize ? (bufsize - 1) : result;
  525. }
  526. void resetError_MBR(void)
  527. {
  528. mbr_errorbuf[0] = 0;
  529. }
  530. #endif // INCLUDE_MBROLA